Repository navigation
Conversation
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
github-actions
Bot
added
vouch:unvouched
ApprovabilityVerdict: Approved at Macroscope's review found this PR approvable — Claude gateway quota failures now support scheduled retries, with web and mobile distinguishing retry times from confirmed quota resets. Recovery remains opt-in, with backoff and existing Stop, cancellation and stale-request safeguards. You can add or adjust custom eligibility rules. Learn more. |
![macroscopeapp[bot]](https://avatars.githubusercontent.com/in/900172?s=60&v=4){kind=small}
|
No actionable comments were generated in the recent review.
 |
| Layer / File(s) | Summary |
|---|---|
Classify Claude quota failures packages/contracts/src/orchestrationV2.ts, packages/provider-core/src/server/failure.ts, apps/server/src/orchestration-v2/Adapters/claudeGatewayQuota.ts, apps/server/src/orchestration-v2/Adapters/ClaudeAdapterV2.ts, apps/server/src/orchestration-v2/Adapters/*test.ts |
The Claude adapter recognizes eligible quota errors and returns retryable usage-limit failures with reset or retry timing. Provider-thread metadata tracks quota retry attempts, and tests cover classification and backoff. |
Propagate retry deadlines through projections packages/shared/src/orchestrationV2ThreadError.ts, apps/server/src/orchestration-v2/ProjectionStore.ts, apps/server/src/orchestration-v2/ProjectionStore.test.ts, packages/client-runtime/src/state/models.ts |
Retry timestamps flow into shell and runtime summaries. Recovery candidate selection uses a retry deadline when the failure is retryable and has no reset time. |
Schedule and continue retry recovery apps/server/src/orchestration-v2/Orchestrator.ts, apps/server/src/orchestration-v2/UsageLimitRecoveryWorker.ts, apps/server/src/orchestration-v2/runtimeLayer.test.ts |
Recovery validation and scheduling use reset or retry timing. Continuations can carry matching auto-resume recovery state to a new run. Runtime tests cover retry deadlines and recovery scenarios. |
Display retry timing and recovery actions apps/web/src/components/ChatView.tsx, apps/web/src/components/chat/UsageLimitRecoveryBanner.tsx, apps/web/src/components/chat/UsageLimitRecoveryBanner.test.tsx, apps/mobile/src/features/threads/UsageLimitRecoveryCard.tsx |
The web banner and mobile card use retry timing when reset timing is absent. Their descriptions and actions distinguish retry times from known reset times. |
Priority:
Estimated code review effort: 4 (Complex) | ~45 minutes
Change: Bug fix
Sequence Diagram(s)
sequenceDiagram
participant ClaudeAdapterV2
participant claudeGatewayQuotaFailure
participant ProjectionStore
participant UsageLimitRecoveryWorker
participant UsageLimitRecoveryBanner
ClaudeAdapterV2->>claudeGatewayQuotaFailure: classify quota API error
claudeGatewayQuotaFailure-->>ClaudeAdapterV2: return retryable failure with retryAt
ClaudeAdapterV2->>ProjectionStore: provide usage-limit failure
ProjectionStore->>UsageLimitRecoveryWorker: expose retry-time recovery candidate
UsageLimitRecoveryWorker->>UsageLimitRecoveryBanner: provide shell retry timing
Suggested labels: bugfix
Suggested reviewers: juliusmarminge, t3dotgg, fixfon
Merge Risk: 
e04a1
-
This change lets Claude gateway quota failures without a known reset time resume automatically through the existing recovery worker. Web and mobile now show the retry time and use retry-specific action labels. No merge-blocking risk was identified in the supplied changes. Interactive client behavior and live provider calls were not exercised.
Pre-merge checks: 
-
Check name Status Explanation Title check The title clearly identifies the main change: resuming Claude turns after gateway quota recovery. It is concise, specific, and uses a conventional commit format. Description check The description covers the problem, implementation, scope boundaries, verification results, limitations, and agent disclosure. It does not use the template headings and does not include the requested … Linked Issues check Check skipped because no linked issues were found for this pull request. Out of Scope Changes check Check skipped because no linked issues were found for this pull request. Full details: Description check
-
Explanation
The description covers the problem, implementation, scope boundaries, verification results, limitations, and agent disclosure. It does not use the template headings and does not include the requested UI screenshots, but it remains sufficiently complete and directly relevant.
-
Finishing Touches
-
🧪 Generate unit tests (beta)
-
- Create a new PR
-
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Comment @coderabbitai help to get the list of available commands.
|
closing at @StiensWout's request under Provide evidence for the changed behavior. Changes the visible web/mobile quota-recovery states to distinguish estimated retry deadlines from confirmed resets, but supplies no before/after screenshots or interaction recording. The PR explicitly says interactive client screenshots were not performed. Its extensive backend recovery tests do not demonstrate those changed client states. Add clear before/after captures of the affected web/mobile recovery states, with a short recording if needed to demonstrate retry timing or interaction. Retain the existing focused backend proof. thanks for trying to help improve t3 code. if you think this closure was a mistake, please open a fresh pr addressing the requirement above, and link back to this one. |
Claude-compatible gateways can return
503 claude_quota_exhaustedor a structured quota error without a known reset time. T3 currently treats these as generic failures or leaves them without a deadline, so the existing automatic-resume workflow never continues them when capacity returns.Recognize the stable gateway quota codes on 429/503, including a result preceded by an assistant rate-limit frame. Keep a real quota reset separate from a persisted retry deadline. When the gateway cannot report a reset or retry hint, use bounded exponential backoff with deterministic jitter. Persist the attempt count in provider-thread metadata and clear it after success.
Reuse the existing durable recovery worker and settings. A per-thread opt-in follows automatic quota continuations through another failed attempt, while Stop, cancellation, pending requests, replacement and manual runs retain their boundaries. Web/desktop and mobile distinguish retries from confirmed resets. No recovery setting or product default changes; this repairs an existing recovery path for an obvious rate-limit failure.
Validation: the integrated branch passed
vp run check:changed origin/main, including 602 tests, affected-package typechecks, formatting, lint and React Compiler checks. Regression tests failed before the fixes. Coverage includes SQL recovery candidates, duplicate delivery, cancellation/Stop races, per-thread opt-in with the global setting off, a later manual run, and a fresh adapter restoring serialized retry metadata through another quota failure and eventual success. Interactive client screenshots and live provider calls were not performed.Written with GPT-6.1 Sol through the Codex harness in T3 Code; independently reviewed with GPT-6 Astra.